NG2 proteoglycan as a pericyte target for anticancer therapy by tumor vessel infarction with retargeted tissue factor

Abstract

tTF-TAA and tTF-LTL are fusion proteins consisting of the extracellular domain of tissue factor (TF) and the peptides TAASGVRSMH and LTLRWVGLMS, respectively. These peptides represent ligands of NG2, a surface proteoglycan expressed on angiogenic pericytes and some tumor cells. Here we have expressed the model compound tTF-NGR, tTF-TAA, and tTF-LTL with different lengths in the TF domain in E. coli and used these fusion proteins for functional studies in anticancer therapy. We aimed to retarget TF to tumor vessels leading to tumor vessel infarction with two barriers of selectivity, a) the leaky endothelial lining in tumor vessels with the target NG2 being expressed on pericytes on the abluminal side of the endothelial cell barrier and b) the preferential expression of NG2 on angiogenic vessels such as in tumors. Chromatography-purified tTF-TAA showed identical Factor X (FX)-activating procoagulatory activity as the model compound tTF-NGR with Km values of approx. 0.15 nM in Michaelis-Menten kinetics. The procoagulatory activity of tTF-LTL varied with the chosen length of the TF part of the fusion protein. Flow cytometry revealed specific binding of tTF-TAA to NG2-expressing pericytes and tumor cells with low affinity and dissociation KD in the high nM range. In vivo and ex vivo fluorescence imaging of tumor xenograft-carrying animals and of the explanted tumors showed reduction of tumor blood flow upon tTF-TAA application. Therapeutic experiments showed a reproducible antitumor activity of tTF-TAA against NG2-expressing A549-tumor xenografts, however, with a rather small therapeutic window (active/toxic dose in mg/kg body weight).

Keywords: NG2 proteoglycan; cancer; truncated tissue factor; vascular infarction; vascular targeting.

Figure 1. Schematic visualization of possible targets close to a tumor vessel wall

Figure 2. Expression of the proteoglycan NG2 in human cell lines and tumor tissue

(A) APAAP staining with an anti-NG2 antibody shows NG2 expression on the surface of human aortic smooth muscle cells (HuAoSMC, b) and human G361-melanoma cells (d). Cells only incubated with the secondary antibody were used as controls (a, c). (B) The APAAP staining of different resected human tumor tissues reveals the expression of the proteoglycan NG2 especially outside of the tumor endothelium in non-small cell lung carcinoma (NSCLC, b), colon carcinoma (d) and melanoma (g), and on the surface of melanoma cells (f). Tissue sections only incubated with the secondary antibody were used as controls (a, c, e). (C) Human tumor xenotransplants in mice (HT1080 fibrosarcoma, upper row; U87 glioblastoma, lower row) were resected and immunostained for NG2 (red; b, e) and for CD31 (green; a, d), respectively. Co-staining with both antibodies proves the co-localization of NG2-expressing pericytes and CD31-expressing endothelial cells (c, f). Nuclei were stained with DAPI (blue).

Figure 3. Schematic illustration of the NG2-targeting tTF-fusion protein constructs and their purification process

(A) The N-terminal His-tag allows the purification of tTF constructs by metalchelate affinity chromatography; the tTF domain consisting of 214 or 218 amino acids, respectively, mediates the coagulation activity; the C-terminal peptide binding motif enables the binding to the tumor endothelium (CD13 via NGR binding motif in the model compound tTF-NGR) or to the tumor pericytes (NG2 via TAA or LTL binding motif), respectively. (B) The four-step purification process of the tTF-fusion proteins comprises an immobilized metal affinity chromatography (IMAC), a gel filtration/buffer exchange step (GF1), an anion exchange chromatography (AIEX), and a final gel filtration/buffer exchange step (GF2). The elution peaks of the particular purification steps are exemplarily shown for tTF₂₁₈-TAA. (C) As an example, purified tTF₂₁₈-TAA fusion protein (molecular weight: ∼30 kDa) and the purification intermediates analyzed by SDS-PAGE and Western blotting with an anti-tTF antibody are shown: (M) molecular weight standard; (1) IMAC flow-through; (2) IMAC eluate; (3) GF1 eluate; (4) AIEX eluate; (5) end product (GF2 eluate). (D) SDS-PAGE analysis of the tTF₂₁₄-LTL purification course. For abbreviations, see above.

Figure 4. Factor X activation assay of the different tTF proteins

(A) The ability of the fusion proteins to enhance the specific proteolytic activation of FX by FVIIa in the presence of phospholipids was evaluated by Michaelis-Menten analysis (a, c). The Michaelis constants (K_m) of the activation were calculated by hyperbolic regression analysis according to Hanes et al. [33] (b, d). (B) The determined K_m values of all tTF constructs are summarized in this scheme. There are no significant differences between the respective short and long constructs as analyzed by two-sided t-test: tTF-NGR: p = 0.85; tTF-TAA: p = 0.97; tTF-LTL: p = 0.056. Data are presented as means +/− standard errors.

Figure 5. Binding of tTF₂₁₈-TAA to human aortic smooth muscle cells (HuAoSMC) as measured by flow cytometry

(A) The presence of NG2 on HuAoSMCs and HUVECs was detected with a monoclonal PE-labeled anti-NG2 antibody. (B) Dose-dependent binding of tTF-TAA to NG2-expressing HuAoSMCs (lower row): cells were incubated with different concentrations of tTF-TAA. Bound fusion protein was then detected with a FITC-labeled anti-His antibody (α His-FITC). tTF without targeting TAA-peptide binds unspecifically to the cells, but in a clearly smaller amount when compared to tTF-TAA (upper row, right panel). Untreated cells (NTC), cells only incubated with the anti-His antibody, or cells incubated both with the control protein BSA and the anti-His antibody, respectively, where used as controls (upper row). A summary of all binding curves is shown in the histogram (lower row, right panel); for color assignment see boxes above the respective panels. (C) Blocking of HuAoSMCs using the PE-labeled anti-NG2 antibody (α NG2-PE) and displacement by tTF-TAA: after incubation with the α NG2-PE antibody, cells were incubated with different concentrations of tTF-TAA. The fusion protein was able to displace some of the bound antibody in a dose-dependent manner (lower left and middle panels). Bound tTF-TAA fusion protein was then detected with the anti-His-FITC antibody (lower right panel). Displacement of the NG2-PE antibody by the control protein BSA was not effective (upper middle panel). (D) In a further setup, tTF-TAA binding was blocked by pre-incubation with 20-fold excess of pure TAASGVRSMH-decapeptide (lower panel). Bound tTF-TAA fusion protein was then detected with the anti-His-FITC antibody. Cells without pre-incubation were used as controls (upper panel).

Figure 6. Binding of tTF₂₁₈-TAA to human G361-melanoma cells as measured by flow cytometry

(A) The presence of NG2 on G361-melanoma cells was detected with a monoclonal PE-labeled anti-NG2 antibody. (B) Dose-dependent binding of tTF-TAA to NG2-expressing G361 cells (lower row): cells were incubated with different concentrations of tTF-TAA. Bound fusion protein was then detected with a FITC-labeled anti-His antibody (α His-FITC). tTF without targeting TAA-peptide binds unspecifically to the cells, but in a clearly smaller amount when compared to tTF-TAA (upper row, right panel). Untreated cells (NTC) or cells only incubated with the anti-His antibody, respectively, where used as controls (upper row, left and middle panel). A summary of all binding curves can be seen in the histogram (lower row, right panel); for color assignment, see boxes above the respective panels. (C) Blocking of G361 with the PE-labeled anti-NG2 antibody (α NG2-PE) and displacement by tTF-TAA: after incubation with α NG2-PE antibody cells were incubated with tTF-TAA, whereupon the fusion protein was able to displace some of the bound antibody (lower left panel). Bound tTF-TAA fusion protein was then detected with the α His-FITC antibody (lower right panel). Displacement of the α NG2-PE antibody by the control protein BSA was not effective (upper right panel). (D, E) In a further setup, tTF-TAA binding was blocked by pre-incubation with 4- to 20-fold excess of pure TAASGVRSMH-dekapeptide in a dose-dependent manner (D, E lower panel). Bound tTF-TAA fusion protein was then detected with the α His-FITC antibody. Untreated cells and cells without pre-incubation were used as controls (upper panel). By increasing the applied concentrations of fusion protein, more already bound TAASGVRSMH could be displaced (E, lower right panel). Blocking of tTF-TAA binding by the control protein BSA was not effective (E, upper right panel).

Figure 7. Determination of the dissociation constant (K_D) of tTF₂₁₈-TAA binding to NG2-expressing HuAoSMC to characterize the binding affinity

Binding assays with different concentrations of radiolabeled tTF-TAA (¹²³I-tTF-TAA) and cultivated human aortic smooth muscle cells (HuAoSMC) were performed and the amount of bound fusion protein was carried out in a Berthold gamma counter. Determination of K_D was performed with a nonlinear curve fitting program (GraphPad Prism 6: one site – total binding) and analysis of total-binding data by means of the equation for (one-site) total binding revealed a K_D of 785 nM. This equation (total binding = {(Bmax * [L]/(K_D+[L])} + NS * [L]) assumes that unspecific binding is commensurate to the concentration of the radioligand and that only a small amount of the radioligand binds, so that the input-concentration is virtually identical with the concentration of free radioligand (for further details see Materials and Methods).

Figure 8. Treatment with NG2-targeting tTF₂₁₈-TAA in vivo and ex vivo as monitored by fluorescence reflectance imaging (FRI)

(A, B) FRI was performed with A549-bearing CD-1 nude mice that received AngioSense^® as an in vivo blood pool- and tumor-imaging agent 22 h prior to the start of the therapy. AngioSense^® signal intensities in tumors were measured at time point zero (0 min; see white arrows) and set as 100%. After the application of 1 mg/kg tTF₂₁₈-TAA or saline, respectively (each with n = 3), signal intensities were analyzed for 120 min and revealed a significant decrease of the signal in treated tumors when compared to the saline controls (B; asterisk denotes statistical significance, p < 0.05). Data are presented as means +/− standard errors. (C, D) AngioSense^® fluorescence intensities of explanted tumors showed an even more significant difference between treated tumors (C, lower row) and saline controls (C, upper row), which was quantified as shown in D (asterisk denotes statistical significance, p < 0.05). Data are presented as means +/− standard errors.

Figure 9. Effect of tTF₂₁₈-TAA on the growth of human xenotransplants in CD-1 nude mice

(A) Tumor growth retardation of human adenocarcinoma of the lung (A549) xenotransplanted into athymic CD-1 mice after i.v. administration of 0.5 (n = 5) and 1 mg/kg bw (n = 4) tTF₂₁₈-TAA versus 0.9% saline control (n = 8). Therapy started at an average tumor size of 240 mm³ and was continued for 16 days on every second day; arrows indicate time points of injection. Due to side effects, the 1 mg/kg therapy cohort had to be stopped at day 10. (B) Effect of 0.5 mg/kg BW tTF₂₁₈-TAA (n = 8), 5 mg/kg bw TMS(PEG)₁₂ tTF-TAA (n = 9) and 0.9% saline control (n = 9), respectively, on the growth of human M21 melanoma xenotransplanted into CD-1 nude mice. Therapy started at an average tumor size of 250 mm³ and was continued for 23 days on every second day; arrows indicate time points of i.v. administration. The experiment had to be terminated with separating growth curves on day 23 due to tumor size and animal regulations. (C, D) Vessel decoration with NG2 of explanted control tumors. Human tumor xenotransplants in mice (×400; C: A549 adenocarcinoma of the lung; D: M21 melanoma) were resected and co-immunostained for NG2 (red) and for CD31 (green), respectively. Nuclei were stained with DAPI (blue).